Researchers at Princeton University are using reinforcement learning and inverse design techniques to create radio-frequency integrated circuits (RFICs) from scratch, producing chip layouts that outperform conventional designs and, in some cases, look nothing like what a human engineer would have built.
RFICs are critical components for 5G networks, autonomous vehicles, satellite communications, and wireless infrastructure. Their design has traditionally been considered an art requiring years of specialized expertise, involving manual iteration through complex electromagnetic trade-offs.
The Princeton team’s approach, developed under a US$10 million grant from the National Semiconductor Technology Center’s AIDRFIC program, uses machine learning algorithms trained on electromagnetic simulations to navigate the enormous design space of RF circuits. The AI generates novel layouts automatically, drastically reducing design time from what typically takes months or years of human effort.
“The project began after the success of AI in games like Go,” the researchers note, drawing a parallel between the combinatorial explosion of possible moves in a board game and the vast number of potential circuit configurations in RF design.
The project, formally named “IMAGINE: Inverse Methods and Generative AI for Algorithmic and Non-intuitive Design Explorations in RFICs,” is led by Kaushik Sengupta, professor of electrical and computer engineering at Princeton. The team includes researchers from the University of Southern California, Drexel University, and Northeastern University, with industry partners including RTX, Keysight, and Cadence. An advisory board features senior leadership from Qualcomm, Skyworks, Texas Instruments, Nokia Bell Labs, Ericsson, and Maury Microwave.
The findings have attracted significant attention within the RF community, sparking discussions about the role of AI in chip design. The researchers emphasize the need for large, shared datasets and open ecosystems to further advance AI-driven electromagnetic and circuit design.
As demand for advanced RFICs grows with the rollout of next-generation wireless infrastructure, the potential for AI-driven design to reshape the field is becoming increasingly apparent. The Princeton team is one of three groups selected by Natcast for this round of AIDRFIC funding; the others are led by Keysight Technologies and the University of Texas at Austin.
Sources: AI Is Designing Radio Chips That Humans Couldn’t Even Imagine (RobotToday, June 24, 2026); Princeton will lead U.S. effort to design better chips for wireless communication (Princeton University, August 4, 2025)